EfficientNet-lite0.md

PyTorch EfficientNet-lite0

Setup AI Model Efficiency Toolkit

Please install and setup AIMET before proceeding further. This model was tested with the torch_gpu variant of AIMET version 1.21.0.

Experiment Setup

Additional dependencies

Install geffnet using pip install

python -m pip install geffnet

Loading AIMET model zoo libraries

export PYTHONPATH=$PYTHONPATH:<aimet_model_zoo_path>

Model checkpoints and configuration

• Downloading checkpoints and Quantization configuration file are handled through evaluation script.
• The original EfficientNet-lite0 checkpoint can be downloaded from here:
  • https://github.com/rwightman/gen-efficientnet-pytorch
• Optimized EfficientNet-lite0 checkpoint can be downloaded from the Releases page.
• The Quantization Simulation (Quantsim) Configuration file can be downloaded from here: default_config_per_channel.json (Please see this page for more information on this file).

Dataset

• This evaluation was designed for the 2012 ImageNet Large Scale Visual Recognition Challenge (ILSVRC2012), which can be obtained from: http://www.image-net.org/ The dataset directory is expected to have 3 subdirectories: train, valid, and test (only the valid test is used, hence if the other subdirectories are missing that is ok). Each of the {train, valid, test} directories is then expected to have 1000 subdirectories, each containing the images from the 1000 classes present in the ILSVRC2012 dataset, such as in the example below:

  train/
  ├── n01440764
  │   ├── n01440764_10026.JPEG
  │   ├── n01440764_10027.JPEG
  │   ├── ......
  ├── ......
  val/
  ├── n01440764
  │   ├── ILSVRC2012_val_00000293.JPEG
  │   ├── ILSVRC2012_val_00002138.JPEG
  │   ├── ......
  ├── ......

Usage

To run evaluation with QuantSim in AIMET, use the following

 python3  efficientnetlite0_quanteval.py \
                --default-param-bw <weight bitwidth for quantization - 8 for INT8, 4 for INT4> \
                --dataset-path < path to validation dataset> \
                --batch-size  <batch size as an integer value> \
                --use-cuda <use GPU or CPU> 

Quantization Configuration

• Weight quantization: 8 or 4 bits per channel symmetric quantization
• Bias parameters are not quantized
• Activation quantization: 8 bits, asymmetric quantization
• Model inputs are quantized
• TF_enhanced was used for weight quantization scheme
• TF was used for activation quantization scheme
• Batch norm folding and Adaround have been applied on optimized efficientnet-lite checkpoint
• [Conv - Relu6] layers has been fused as one operation via manual configurations
• 4K Images from ImageNet training dataset (4 images per class) are used as calibration dataset
• Standard ImageNet validation dataset are usef as evaluation dataset